Every day they surprise us with the fact that they allow a better user experience. Today’s news has to do with the Facebook image search engine. Now the social network allows you to search for images by describing their content Thanks to the .
was conceived and launched to improve the experience of visually impaired users in the Facebook community, but now the computer vision platform, Lumos, is driving Facebook’s image search engine be available to all users.
The Facebook image search engine allows you to search by description
Hashtags or tags made it easier to track some events and classify images, but there are times when even that doesn’t help us find the photo we’re looking for. This is where the rescue comes in, since now simply By describing the content of the image, the algorithm will be able to interpret the terms you enter through the search engine.
To accomplish this task, Facebook created Lumos, a neural network with tens of millions of images. Facebook hosts billions of images that users have shared at some point, which makes this task easy.
you only have to capture the description of the photograph that you are interested in locating in the search engine and then select the tab “images” in the results, so that it shows you your photos, those of your contacts, and the public photos of all the users that fit those search criteria.
The system will be able to review each photo and find the matches with the established search criteria. After matching the terms to the images, Lumos classifies its output using information from the images and the original search.
Lumos is also applied to help the visually impaired on Facebook
Along with Facebook’s new image search engine, the This tool works for visually impaired users who can take advantage of existing text-to-speech tools and be able to understand the content of the photos.
The problem was that the elements that made up the photograph were mentioned as “dance, stage, theater” but it was not very clear. Now Facebook has solved that problem by training a computer vision model to identify the actions that happen in the photos. Now you can hear the description like “people dancing on a stage”, a better description and much better contextualized.
You can apply the same in the new Facebook image search engine where you can perform the search by the actions that happen in the imagessuch as “people playing instruments”.
How the Lumos Image Search System Works
“A large percentage of photos shared on Facebook include people, so we focused on provide automated descriptions involving people”. The Artificial Intelligence team gathered a sample of 130,000 photos shared on Facebook that included individuals”, , Joaquin Quinonero Candela.
“They were asked to write a one line description of the photo, as if they were describing it to a visually impaired friend. We then use these annotations to build a learning model automatic that can continuously discover people’s actions“, Explain Quinonero.
Eventually, Facebook will apply this technology to your . This type of search can be used both to find the exact moment when your friend blew out the candles on his cake, as in a commercial context. And of course this last context may increase them in the future.
At the moment, this function only available in the United States and the results will only be able to show the photos that are marked as public on the social network, but it is expected to expand globally in the coming weeks.
conclusion
Artificial intelligence is here to stay, and Facebook cannot ignore this reality. For this reason, the facebook image search it is a very useful tool. You may need an image of a certain content and, thanks to Lumos, you can easily find it. This technology has several applications in the professional field that, we trust, will be especially fruitful. Integration with Facebook will allow you to achieve more with less. As against, it should be noted that its implementation is being slower than we would like. It is missing that this new application takes advantage of all the potential that it has.
Image:
Stay informed of the most relevant news on our news channel
Facebook Comments
Disqus